Quality Water Index Polynomial (QWIP)

Quality Water Index Polynomial (QWIP)

Table of Contents

  1. Product Summary
  2. Algorithm Description
  3. Implementation
  4. Assessment
  5. References
  6. Data Access

1 - Product Summary

This algorithm returns the Quality Water Index Polynomial (QWIP) score, representing a quantitative metric to evaluate the quality of ocean color remote sensing reflectance ($R_{rs}$) data (Dierssen et al. 2022). The relationship between the Apparent Visible Wavelength (AVW; Vandermeulen et al. 2020) and a multi-channel waveband index is used to identify spectra that fall outside the general trends observed in aquatic optics for optically deep waters. The approach was developed with a large global dataset representing blue, green, and brown waters and was further tested extensively with field and satellite datasets. This simple approach can provide a level of uncertainty about a retrieved spectrum and flag questionable or unusual spectra for further analysis.

The color-coded output of the QWIP algorithm identifies the location of unusual spectral shape features. This screening procedure provides a robust and automated quality control mechanism across different water types, from blue open ocean waters, to dense freshwater algae blooms or sediment-dominated coastal waters.

Algorithm Point of Contact: Heidi Dierssen, University of Connecticut

2 - Algorithm Description

Inputs: $R_{rs}$ at all available wavelengths between 400 – 700 nm (rrs_vvv).

Outputs: qwip, Quality Water Index Polynomial score (unitless)

Approach: The Quality Water Index Polynomial (QWIP) is a mathematical model relating the calibrated Apparent Visible Wavelength (AVW, Vandermeulen et al. 2020) to a normalized difference index (NDI). The QWIP score is calculated as the difference between a measured and AVW-predicted NDI. To initiate, the measured NDI is determined as:

$$NDI = \frac{R_{rs} (\lambda_2) - R_{rs} (\lambda_1)}{R_{rs}(\lambda_2) + R_{rs} (\lambda_1)}$$

where $\lambda1$ = Rrs_490, and $\lambda2$ = Rrs_665 (for multispectral sensors, the closest wavelength match is used).

The predicted NDI is related to the Apparent Visible Wavelength (AVW) as follows:

$$NDI_{predicted}=p_1AVW^{4}+p_2 AVW^{3}+p_3 AVW^{2}+p_4 AVW+p_5$$ $$p=(-8.399885\times10^{-9},1.715532 \times10^{-5},-1.301670 \times 10^{-2},4.357838 \times 10^{0},-5.449532 \times 10^{2} )$$

Finally, the QWIP score is calculated as the difference between the NDI and NDI_predicted:


The NDI provides a means to highlight the variability of logarithmically distributed data on a linear scale such that the distance either above or below the central tendency (QWIP) can be scored with a positive or negative value. Generally, hyperspectral data with QWIP scores exceeding a value of |0.2| may be subject to additional screening to determine any evident spectral anomalies. It maybe necessary to relax the nominal threshold (e.g. |0.3|) when applying QWIP to multispectral sensors.

3 - Implementation

Product Short Name: qwip

Level-2 Product Suite: None

Calling in L2GEN: l2prod = qwip

qwip_coef = [p1,p2,p3,p4,p5]

4 - Assessment

Algorithm Development: The method was developed using a large global dataset of remote sensing reflectance (n = 1,629) compiled from different studies (CASCK-P dataset, see Dierssen et al. 2022).

(A) The QWIP relationship between Apparent Visible Wavelength (AVW) and the Normalized Difference Index (NDI) with the CASCK-P training dataset showing the final tuned QWIP polynomial (thick magenta line) with different levels of QWIP scores (±0.1 dotted magenta and ±0.2 dashed magenta). Water types include: Blue-green (blue circles), Green (green diamonds) and Brown (red squares). (B) Histogram of the QWIP scores from (A) are predominantly within ±0.1 for the training data. (C) The remote sensing reflectance ($R_{rs}$) of outliers with negative QWIP scores < −0.2 were associated with optically shallow water features. (D) Outliers with QWIP scores > 0.2 exhibited higher blue associated with surface reflected skylight and higher overall magnitude spectra.

Algorithm Verification: The QWIP approach was tested using several different regional field datasets collected with above-water methodology and on satellite-retrievals of water-leaving reflectance data.

In situ verification:

(A) The QWIP scores from highly quality controlled hyperspectral PANTHYR reflectance data from Vanhellemont (2020). (B) QWIP scores were predominantly within <± 0.1. Water types include: Blue-green (blue circles), Green (green diamonds) and Brown (red squares).

In situ verification:

(A) The QWIP approach was used for quality control of a raw WISP dataset. The red circle highlights false positive data of brown water type that coincidentally fall within the polynomial limits. (B) The majority of the data had QWIP scores of ±0.2. (C) Remote sensing reflectance ($R_{rs}$) of green outliers with slightly negative scores had good spectral shapes but too low in the blue. (D) High QWIP scores were related to the outliers of bad data with unusual spectral shapes. (E) Brown outliers with failing spectral shapes were identified as having lower AVW than expected for the water type (AVW < 540 nm).

Satellite verification:

Comparing full spectral information against empirical indices enables a quick and efficient means of assessing the relative quality of satellite and/or in situ data. Mapped HICO scenes are passed through the QWIP procedure, and spectra are accepted/rejected based on a nominal acceptance threshold. As spectral data increasingly deviate from the polynomial relationship between AVW and NDI (490,665), the anomalous spectral features become more prominent.

5 - References

Dierssen, H. M., Vandermeulen, R. A., Barnes, B. B., Castagna, A., Knaeps, E., & Vanhellemont, Q., 2022: “QWIP: A Quantitative Metric for Quality Control of Aquatic Reflectance Spectral Shape using the Apparent Visible Wavelength,” Frontiers in Remote Sensing, 32. https://doi.org/10.3389/frsen.2022.869611

Vandermeulen, R. A., Mannino, A., Craig, S.E., Werdell, P.J., 2020: “150 shades of green: Using the full spectrum of remote sensing reflectance to elucidate color shifts in the ocean,” Remote Sensing of Environment, 247, 111900, https://doi.org/10.1016/j.rse.2020.111900

Vanhellemont, Q., 2020: “Sensitivity analysis of the dark spectrum fitting atmospheric correction for metre-and decametre-scale satellite imagery using autonomous hyperspectral radiometry,” Optics Express, 28(20), 29948-29965, https://doi.org/10.1364/OE.397456

6 -Data Access